Commit Graph

19 Commits

Author SHA1 Message Date
Pedro F. Giffuni
6244c6e7db sys/rpc: minor spelling fixes.
No functional change.
2016-05-06 01:49:46 +00:00
Garrett Wollman
3c42b5bf28 Fix overflow bugs in and remove obsolete limit from kernel RPC
implementation.

The kernel RPC code, which is responsible for the low-level scheduling
of incoming NFS requests, contains a throttling mechanism that
prevents too much kernel memory from being tied up by NFS requests
that are being serviced.  When the throttle is engaged, the RPC layer
stops servicing incoming NFS sockets, resulting ultimately in
backpressure on the clients (if they're using TCP).  However, this is
a very heavy-handed mechanism as it prevents all clients from making
any requests, regardless of how heavy or light they are.  (Thus, when
engaged, the throttle often prevents clients from even mounting the
filesystem.)  The throttle mechanism applies specifically to requests
that have been received by the RPC layer (from a TCP or UDP socket)
and are queued waiting to be serviced by one of the nfsd threads; it
does not limit the amount of backlog in the socket buffers.

The original implementation limited the total bytes of queued requests
to the minimum of a quarter of (nmbclusters * MCLBYTES) and 45 MiB.
The former limit seems reasonable, since requests queued in the socket
buffers and replies being constructed to the requests in progress will
all require some amount of network memory, but the 45 MiB limit is
plainly ridiculous for modern memory sizes: when running 256 service
threads on a busy server, 45 MiB would result in just a single
maximum-sized NFS3PROC_WRITE queued per thread before throttling.

Removing this limit exposed integer-overflow bugs in the original
computation, and related bugs in the routines that actually account
for the amount of traffic enqueued for service threads.  The old
implementation also attempted to reduce accounting overhead by
batching updates until each queue is fully drained, but this is prone
to livelock, resulting in repeated accumulate-throttle-drain cycles on
a busy server.  Various data types are changed to long or unsigned
long; explicit 64-bit types are not used due to the unavailability of
64-bit atomics on many 32-bit platforms, but those platforms also
cannot support nmbclusters large enough to cause overflow.

This code (in a 10.1 kernel) is presently running on production NFS
servers at CSAIL.

Summary of this revision:
* Removes 45 MiB limit on requests queued for nfsd service threads
* Fixes integer-overflow and signedness bugs
* Avoids unnecessary throttling by not deferring accounting for
  completed requests

Differential Revision:	https://reviews.freebsd.org/D2165
Reviewed by:	rmacklem, mav
MFC after:	30 days
Relnotes:	yes
Sponsored by:	MIT Computer Science & Artificial Intelligence Laboratory
2015-04-01 00:45:47 +00:00
Rick Macklem
c59e4cc34d Merge the NFSv4.1 server code in projects/nfsv4.1-server over
into head. The code is not believed to have any effect
on the semantics of non-NFSv4.1 server behaviour.
It is a rather large merge, but I am hoping that there will
not be any regressions for the NFS server.

MFC after:	1 month
2014-07-01 20:47:16 +00:00
Alexander Motin
b563304c50 Split RPC pool threads into number of smaller semi-isolated groups.
Old design with unified thread pool was good from the point of thread
utilization.  But single pool-wide mutex became huge congestion point
for systems with many CPUs.  To reduce the congestion create several
thread groups within a pool (one group for every 6 CPUs and 12 threads),
each group with own mutex.  Each connection during its registration is
assigned to one of the groups in round-robin fashion.  File affinify
code may still move requests between the groups, but otherwise groups
are self-contained.

MFC after:	2 weeks
Sponsored by:	iXsystems, Inc.
2014-06-08 11:19:32 +00:00
Alexander Motin
b5d7fb7398 Remove st_idle variable, duplicating st_xprt.
MFC after:	2 weeks
2014-06-08 10:18:22 +00:00
Alexander Motin
b776fb2d67 Introduce new per-thread lock to protect the list of requests.
This allows to slightly simplify svc_run_internal() code: if we processed
all the requests in a queue, then we know that new one will not appear.

MFC after:	2 weeks
2014-06-08 09:40:26 +00:00
Peter Wemm
bcea84bd86 Don't expose svc_loss_reg / _unreg to userland as they're kernel-only
additions from r260229 and the SVCPOOL type doesn't exist in userland.
2014-01-08 22:37:18 +00:00
Alexander Motin
0979970a1d Fix NULL dereference panic on UDP requests introduced in r260229. 2014-01-06 12:40:46 +00:00
Alexander Motin
c809a67a72 Replace locks added in r260229 to protect sequence counters with atomics.
New algorithm does not create additional lock congestion, while some races
it includes should not be a problem.  Those races may keep requests in DRC
cache for some more time by returning ACK position smaller then actual,
but it still should be able to drop thems when proper ACK finally read.

Races of the original algorithm based on TCP seq number were worse because
they happened when reply sequence number were recorded. After that even
correctly read ACKs could not clean DRC sometimes.
2014-01-04 15:51:31 +00:00
Alexander Motin
d473bac729 Rework NFS Duplicate Request Cache cleanup logic.
- Introduce additional hash to group requests by hash of sockref.  This
allows to process TCP acknowledgements without looping though all the cache,
and as result allows to do it every time.
 - Indroduce additional callbacks to notify application layer about sockets
disconnection.  Without this last few requests processed just before socket
disconnection never processed their ACKs and stuck in cache for many hours.
 - Implement transport-specific method for tracking reply acknowledgements.
New implementation does not cross multiple stack layers to get the data and
does not have race conditions that previously made some requests stuck
in cache.  This could be done more efficiently at sockbuf layer, but that
would broke some KBIs, while I don't know other consumers for it aside NFS.
 - Instead of traversing all DRC twice per request, run cleaning only once
per request, and except in some conditions traverse only single hash slot
at a time.

Together this limits NFS DRC growth only to situations of real connectivity
problems.  If network is working well, and so all replies are acknowledged,
cache remains almost empty even after hours of heavy load.  Without this
change on the same test cache was growing to many thousand requests even
with perfectly working local network.

As another result this reduces CPU time spent on the DRC handling during
SPEC NFS benchmark from about 10% to 0.5%.

Sponsored by:	iXsystems, Inc.
2014-01-03 15:09:59 +00:00
Alexander Motin
f8fb069d47 Move most of NFS file handle affinity code out of the heavily congested
global RPC thread pool lock and protect it with own set of locks.

On synthetic benchmarks this improves peak NFS request rate by 40%.
2013-12-30 20:23:15 +00:00
Alexander Motin
5c42b9dc1f Introduce xprt_inactive_self() -- variant for use when sure that port
is assigned to thread.  For example, withing receive handlers.  In that
case the function reduces to single assignment and can avoid locking.
2013-12-29 11:19:09 +00:00
Alexander Motin
ba981145d6 Remove several linear list traversals per request from RPC server code.
Do not insert active ports into pool->sp_active list if they are success-
fully assigned to some thread.  This makes that list include only ports that
really require attention, and so traversal can be reduced to simple taking
the first one.

  Remove idle thread from pool->sp_idlethreads list when assigning some
work (port of requests) to it.  That again makes possible to replace list
traversals with simple taking the first element.
2013-12-20 17:39:07 +00:00
Hiroki Sato
2e322d3796 Replace Sun RPC license in TI-RPC library with a 3-clause BSD license,
with the explicit permission of Sun Microsystems in 2009.
2013-11-25 19:04:36 +00:00
Rick Macklem
e2adc47dbb Add support for backchannels to the kernel RPC. Backchannels
are used by NFSv4.1 for callbacks. A backchannel is a connection
established by the client, but used for RPCs done by the server
on the client (callbacks). As a result, this patch mixes some
client side calls in the server side and vice versa. Some
definitions in the .c files were extracted out into a file called
krpc.h, so that they could be included in multiple .c files.
This code has been in projects/nfsv4.1-client for some time.
Although no one has given it a formal review, I believe kib@
has taken a look at it.
2012-12-08 00:29:16 +00:00
Rick Macklem
a4fa5e6dd9 Fix two races in the server side krpc w.r.t upcalls:
Add a flag so that soupcall_clear() is only called once to cancel
  an upcall.
  Move the test for xprt_registered in the upcall down to after the
  mtx_lock() of the pool mutex, to catch the case where it is
  unregistered while the upcall is waiting for the mutex.
Also, move the mtx_destroy() of the pool mutex to after SVC_RELEASE(),
so that it isn't destroyed before the upcalls are disabled.

Reviewed by:	dfr, jhb
Tested by:	pho
Approved by:	kib (mentor)
2009-06-04 14:13:06 +00:00
Rick Macklem
201e7488b6 Added a field to the SVCXPRT structure that the nfsv4 server can
use to identify if the socket is the same one that a cached request
	came in on. It is set by nfsrvd_addsock() to a unique value generated
	by incrementing an unsigned 64bit static variable for each assignment
	and then the value of xp_sockref is tested to see if it is equal to
	the value that was saved with the cached reply.

Submitted by:	rmacklem
Reviewed by:	dfr
Approved by:	kib (mentor)
2009-04-16 16:26:35 +00:00
Doug Rabson
a9148abd9d Implement support for RPCSEC_GSS authentication to both the NFS client
and server. This replaces the RPC implementation of the NFS client and
server with the newer RPC implementation originally developed
(actually ported from the userland sunrpc code) to support the NFS
Lock Manager.  I have tested this code extensively and I believe it is
stable and that performance is at least equal to the legacy RPC
implementation.

The NFS code currently contains support for both the new RPC
implementation and the older legacy implementation inherited from the
original NFS codebase. The default is to use the new implementation -
add the NFS_LEGACYRPC option to fall back to the old code. When I
merge this support back to RELENG_7, I will probably change this so
that users have to 'opt in' to get the new code.

To use RPCSEC_GSS on either client or server, you must build a kernel
which includes the KGSSAPI option and the crypto device. On the
userland side, you must build at least a new libc, mountd, mount_nfs
and gssd. You must install new versions of /etc/rc.d/gssd and
/etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf.

As long as gssd is running, you should be able to mount an NFS
filesystem from a server that requires RPCSEC_GSS authentication. The
mount itself can happen without any kerberos credentials but all
access to the filesystem will be denied unless the accessing user has
a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There
is currently no support for situations where the ticket file is in a
different place, such as when the user logged in via SSH and has
delegated credentials from that login. This restriction is also
present in Solaris and Linux. In theory, we could improve this in
future, possibly using Brooks Davis' implementation of variant
symlinks.

Supporting RPCSEC_GSS on a server is nearly as simple. You must create
service creds for the server in the form 'nfs/<fqdn>@<REALM>' and
install them in /etc/krb5.keytab. The standard heimdal utility ktutil
makes this fairly easy. After the service creds have been created, you
can add a '-sec=krb5' option to /etc/exports and restart both mountd
and nfsd.

The only other difference an administrator should notice is that nfsd
doesn't fork to create service threads any more. In normal operation,
there will be two nfsd processes, one in userland waiting for TCP
connections and one in the kernel handling requests. The latter
process will create as many kthreads as required - these should be
visible via 'top -H'. The code has some support for varying the number
of service threads according to load but initially at least, nfsd uses
a fixed number of threads according to the value supplied to its '-n'
option.

Sponsored by:	Isilon Systems
MFC after:	1 month
2008-11-03 10:38:00 +00:00
Doug Rabson
dfdcada31e Add the new kernel-mode NFS Lock Manager. To use it instead of the
user-mode lock manager, build a kernel with the NFSLOCKD option and
add '-k' to 'rpc_lockd_flags' in rc.conf.

Highlights include:

* Thread-safe kernel RPC client - many threads can use the same RPC
  client handle safely with replies being de-multiplexed at the socket
  upcall (typically driven directly by the NIC interrupt) and handed
  off to whichever thread matches the reply. For UDP sockets, many RPC
  clients can share the same socket. This allows the use of a single
  privileged UDP port number to talk to an arbitrary number of remote
  hosts.

* Single-threaded kernel RPC server. Adding support for multi-threaded
  server would be relatively straightforward and would follow
  approximately the Solaris KPI. A single thread should be sufficient
  for the NLM since it should rarely block in normal operation.

* Kernel mode NLM server supporting cancel requests and granted
  callbacks. I've tested the NLM server reasonably extensively - it
  passes both my own tests and the NFS Connectathon locking tests
  running on Solaris, Mac OS X and Ubuntu Linux.

* Userland NLM client supported. While the NLM server doesn't have
  support for the local NFS client's locking needs, it does have to
  field async replies and granted callbacks from remote NLMs that the
  local client has contacted. We relay these replies to the userland
  rpc.lockd over a local domain RPC socket.

* Robust deadlock detection for the local lock manager. In particular
  it will detect deadlocks caused by a lock request that covers more
  than one blocking request. As required by the NLM protocol, all
  deadlock detection happens synchronously - a user is guaranteed that
  if a lock request isn't rejected immediately, the lock will
  eventually be granted. The old system allowed for a 'deferred
  deadlock' condition where a blocked lock request could wake up and
  find that some other deadlock-causing lock owner had beaten them to
  the lock.

* Since both local and remote locks are managed by the same kernel
  locking code, local and remote processes can safely use file locks
  for mutual exclusion. Local processes have no fairness advantage
  compared to remote processes when contending to lock a region that
  has just been unlocked - the local lock manager enforces a strict
  first-come first-served model for both local and remote lockers.

Sponsored by:	Isilon Systems
PR:		95247 107555 115524 116679
MFC after:	2 weeks
2008-03-26 15:23:12 +00:00